What's up? You're reading Inclined AI. Twitter broke yesterday, so I finally got around to filing my taxes.

Here's what is on tap today:

• AI models scaling & GPT-4

• The newsroom’s newest team member: AI

• Microsoft adds more AI tools to their Business app suite

• AI is the king of aerial combat. Watch out, Maverick

Be a sponsor of our newsletter

CRITICS ARE STARTING TO COME AFTER GPT-4

The model isn’t even out yet, but the criticisms are sprouting up like weeds in our beautiful AI garden.

Jacob Stern, a staff writer for The Atlantic, wrote a piece yesterday arguing that GPT-4 might end up as a bloated, pointless mess.

Ouch.

He takes aim at AI “evangelists” for overhyping this new model and calling it the next best thing without proof. We’ve covered this subject in a recent post, so I wanted to drill down further.

Is scaling the solution we all think it is?

When researchers want to improve AI, they add parameters and scale the amount of ingested data. In machine learning, bigger is better.

At least, that’s what people tell me.

Stern argues that we’ve reached max cramming for GPT and that this next model will get less bang for its buck. I agree; we don’t need 100 trillion parameters to improve something like ChatGPT.

Right now, that’s the only proven way to iterate on a concept. You make it bigger, and you teach it some more. That’s the end of the story.

Or is it?

GPT-4 will flip this script, and even the folks at DeepMind know that scaling isn’t the end-all, be-all for growth. 

How will this happen? OpenAI is spending more time on its learning techniques and customizability for the next phase. So they won’t bloat the concept and provide a clear point to the process.

You can criticize the concept of scaling, but that doesn’t make GPT-4 inherently pointless.

GPT-4 would have been out last year if they only wanted to add parameters.

WANT TO KNOW MORE ABOUT GPT-4?

We did a full write-up of what we know about the new model on our premium Substack.

We have a special deal for you if you want to read it now.

ARE JOURNALISTS AFRAID OF AI?

I assume the answer is yes, given the huge CNET layoffs after they announced they used AI to write articles.

Yet, I am a journalist, and I am not afraid.

No, really, I went to school for Journalism. 

I know anyone can call themselves a journalist now, but I legit studied the history and principles of this craft.

I see these changes firsthand.

AI will disrupt this industry, and I have no doubt it will anger people.

But I want to share a recent article published by Nieman Lab. 

In it, Jeff Israely compares the adoption of generative AI in the newsroom to the use of Google Translate in its early days. Or, going even further back, the growing use of Google in 2002.

Do you remember what news was like pre-2002?

Your weather, your plans, your diet, and your sports scores all game from a journalistic source. No one gave a crap about SEO or algorithms.

The TL;DR of Israely’s post is that journalists shunned Google, before eventually accepting the disruption.

I expect AI to disrupt the flow of information at a scale 10x that of Google in the early 2000s. But it’s safe to assume journalists will adapt and find new ways to use these tools in their reporting again.

For instance, I am beyond excited to see what AI can do to help investigative reporting. More publishers can afford investigative teams when the person writing the story doesn’t need to comb through decades of confidential data.

Also, did we forget what AI learns from?

Circling back around, AI consumes media written by someone somewhere. Journalists are the chefs that give Large Language Models five-star meals.

The concept of a journalist will only loosen with time, and firsthand reporting will become more valuable than ever.

Don’t let automation scare you.

We will all create new things to value, trust me. Or embrace Singularity, whichever works.

Quick Nuggets

🇺🇸 The US House subcommittee for cybersecurity is tackling AI this week

🧑‍💻 Copilot is being added to Microsoft’s full range of business suite apps

❤️ AI love: nah, they’re just not that into us, you guys.

💥 AI manga sparks a huge debate over the validity of the artistic value in generative ai

🛩️ Get schooled: an AI pilot trumped its human counterpart in an aerial dogfight

⛑️ AI in medicine: doctors are using AI systems to help identify new drug candidates who have Parkinson’s Disease

3️⃣ AI, deepfakes, and Bitcoin SV is one weird ass trio to draw a connection with; that’s for sure

💰 $601 million in funding went to Abound so that they can scale their AI-based consumer lending platform

🤔 What’s the deal? Here’s why reinforcement learning is so popular with LLMs like ChatGPT

🚫 Yikes. Motherboard, a VICE tech team, asked ChatGPT for BDSM scenarios, and the result was terrible

💧 The firehose that is AI media can feel uncontrollable, right?

🔥 Fresh Products

• Microsoft Designer - it’s their AI-driven Canva-like product (link)

• Realm - an app that’s AI-assisted social networking (link)

• Scribble Diffusion - is back and makes it even easier to turn a sketch into a work of art (link)

• Kay - a chat AI that lives within WhatsApp (link)

• ChatSpot - like OpenAI and Hubspot had a baby (link)

• Q - a slack bot powered by GPT-3.5 API (link)

• MessengerX - now allows you to talk with Satoshi Nakamoto about Bitcoin (link)

• Studio M64 - made a GPT-3.5 playground to test out the new API (link)

• Seashore.AI - they aim to build the best customer-facing chat AI out there (link)

Good Content, 24/7 Interviews

A new AI Twitch stream is out. This one allows viewers to ask AI-versions of celebrities any question and hear the generated answers. The era of AI entertainment is here.

_{That’s it for today. I hope you enjoyed the latest edition of inclined.ai - Davis.}

Want more?

Upgrade today for less than the price of a latte a month and get benefits like:

• A weekly Sunday deep dive into AI trends

• Access to Q&A privileges to get your biggest AI questions answered

• Exclusive discounts & invites to beta AI tools

Click down below to support our writing 🙂